What's new

My crack at a Mitch Hawker style poll

balrog

Mega Poster
The downside to cross-wise comparison is larger ballots get greater favoritism to larger ballots. That is, if you have ridden more roller coasters, you carry greater sway in how each roller coaster performs in the comparison.
.
This is one of the reasons why, instead of a simple but not really relevant "percentage of win" score, you need to use a ELO (chess) rating or similar on your poll results. This is used in almost every competitive sport/game ranking and for a good reason : it works better. It does not give advantage to the coasters with a higher number of votes, it does not give extra importance to larger samples, it is just fair, plain and simple, as long as the ridership of a ride is not critically low. Plus, the math are really simple and easy to implement.
 

JJLehto

Hyper Poster
The ELO is a great idea. I want to try and remove quantity bias. And I guess nothing can stop blatant abuse, but a good enough poll still works. We all debate specifics but few question the overall MH rankings and agree its far better than Golden Ticket
 

Edward M

Strata Poster
I might have, say, 50 to 100 roller coasters to choose from. Let people put in their personal order preference of what they've ridden. Then, they put in their coaster count which can give people with more coasters more credit and weight in their opinions. It could be where you simply take the premade list and sort them out.

It may not be to the lengths of Hawker or really be that unsusceptible to people's certain preferences. However, I feel it would be (relatively) simple to organize and do. Just an idea.
 

JJLehto

Hyper Poster
I might have, say, 50 to 100 roller coasters to choose from. Let people put in their personal order preference of what they've ridden. Then, they put in their coaster count which can give people with more coasters more credit and weight in their opinions. It could be where you simply take the premade list and sort them out.

It may not be to the lengths of Hawker or really be that unsusceptible to people's certain preferences. However, I feel it would be (relatively) simple to organize and do. Just an idea.
A good idea! I am open to suggestions on having it be less daunting. The drag and drop application was very easy, as it was just a list and you can easily drag and move the rides. Which is why I am looking for potential tech help/tips. Otherwise I figured Id make a spread sheet you fill in, and use coaster count or a full coaster list to go from
 

balrog

Mega Poster
Or, you ask people to give a text file with list of rcdb numbers, and get some simple code to crunch the data with ELO rating or else, and then give you directly a text file with the ranking. Simple, effective, less than an hour of work.
 

Coaster Hipster

Giga Poster
Seeing a MH-style poll come back would be very exciting! Agree about a 50 or 100 coaster cap. Sorting out my Top 20 is fairly managable and interesting to do, deciding which one is #156 or #158 feels just tedious and irrelevant. As JJLehto said in his intro, with all the new RMC, Intamin, Mack and even Vekoma beauties built since the last MH poll, would be fascinating to see the results of a new survey :)
 

balrog

Mega Poster
Well, if you go with Elo-like rating instead of a simple percentage of wins, you don't need a cap. People can send a top 10 or a top 300 if they feel like it. Probably the best way to maximise the number of participants.
 

Hyde

Matt SR
Staff member
Moderator
Social Media Team
ELO and MH actually share the same basic methodology, as both are pairwise designs, meaning they compare the quantitative value of two entities, which infers preference on which is better. ELO actually takes things further, as the ELO score is a predictive rating of how likely one chess player/sports team is to beat their opponent.

Creating an ELO rating for coasters though would be a neat idea, as that would not only give a basis for ranking #1, #2, etc., but also give a sense of how strongly preferred each roller coaster is.
 

balrog

Mega Poster
ELO and MH actually share the same basic methodology, as both are pairwise designs, meaning they compare the quantitative value of two entities, which infers preference on which is better. ELO actually takes things further, as the ELO score is a predictive rating of how likely one chess player/sports team is to beat their opponent.

Creating an ELO rating for coasters though would be a neat idea, as that would not only give a basis for ranking #1, #2, etc., but also give a sense of how strongly preferred each roller coaster is.


Not just that, contrarily to most simple rating system, ELO does not favour players with a higher number of games. Moreover , a strong player do not earn much points by defeating a weak one. The first of the previous characteristics allows to avoid the main problem with Golden Ticket, while the later avoids the main problem with MH ranking.
If it is of any help, I'm willing to write a program that does the ranking when provided the top N as lists of rcdb numbers. I actually wrote it on a sheet of paper last friday while I was bored to death teaching some practical class. I just need to copy it on a computer.
 

HeartlineCoaster

Theme Park Superhero
Any progress on this?

Regarding some of the above comments, the disadvantage of a cap is you're trying to get as many mutual comparisons as possible. You'll miss out on some important statistics if you don't know every ride each person is comparing against. Rides that are love/hate will fare better because they may not feature at all on some top 50s but not be knocked down for it.

Always seemed best to do a 'top however-many-you-want' and then lump the rest of the count into brackets for easy comparison.
If people were willing to submit those, it would be fairly easy to produce some results.
 

Hyde

Matt SR
Staff member
Moderator
Social Media Team
^ There's a trade-off on ease of use however. Folks with 200+ ride counts have often not participated in Mitch Hawker polls of old, as it simply was too time intensive. Having a 50 coaster cap means all ballots are the same length, and could conveniently be finished in minutes. You run the risk of "Golden Ticket-ism" which favors roller coasters that appear on more brackets vs. actual rankings; however spanning a bracket out to 50 relatively mitigates this factor.
 

HeartlineCoaster

Theme Park Superhero
It's not that much more difficult if you're doing a top 50 to then just list everything else below that with a simple good/bad/vile type category.
It makes sure all comparisons are represented. Coaster A may get the best results out of appearing in top 50s, but there could be many people who put it completely outside of that.
 
Top